Figma REST API rate limits differ by endpoint cost

Designing an integration that reads Figma files and renders images via the REST API. The goal is to understand how rate limiting applies across endpoint types so throughput can be planned without assuming uniform limits.

Documentation indicates limits are enforced per endpoint and per token, with expensive operations such as image rendering and file export subject to stricter constraints than lightweight metadata reads. Plugins and widgets use the Plugin API rather than REST, so their constraints are separate.

Current uncertainty is about how limits are communicated and whether pagination patterns affect consumption. Exact numeric thresholds are not fully published and may change, and guidance around retry behavior is policy-sensitive.

What signals does the API provide when a limit is reached for a given endpoint? Is rate limiting applied strictly per user or token, and does distributing requests across multiple tokens for the same user change effective throughput? What pagination practices avoid amplifying limit consumption on list endpoints?